Marine Habitat Mapping Using Bathymetric LiDAR Data: A Case Study from Bonne Bay, Newfoundland

Marine habitats provide various benefits to the environment and humans. In this regard, an accurate marine habitat map is an important component of effective marine management. Newfoundland’s coastal area is covered by different marine habitats, which should be correctly mapped using advanced techno...

Full description

Bibliographic Details
Published in:Water
Main Authors: Meisam Amani, Candace Macdonald, Abbas Salehi, Sahel Mahdavi, Mardi Gullage
Format: Article in Journal/Newspaper
Language:English
Published: MDPI AG 2022
Subjects:
marine habitats
aquatic vegetation
LiDAR
remote sensing
image classification
Hydraulic engineering
TC1-978
Water supply for domestic and industrial purposes
TD201-500
Newfoundland
Online Access:https://doi.org/10.3390/w14233809
https://doaj.org/article/fcc87b75e9ed414992650b0cb65524bc
Description
Summary:Marine habitats provide various benefits to the environment and humans. In this regard, an accurate marine habitat map is an important component of effective marine management. Newfoundland’s coastal area is covered by different marine habitats, which should be correctly mapped using advanced technologies, such as remote sensing methods. In this study, bathymetric Light Detection and Ranging (LiDAR) data were applied to accurately discriminate different habitat types in Bonne Bay, Newfoundland. To this end, the LiDAR intensity image was employed along with an object-based Random Forest (RF) algorithm. Two types of habitat classifications were produced: a two-class map (i.e., Vegetation and Non-Vegetation) and a five-class map (i.e., Eelgrass, Macroalgae, Rockweed, Fine Sediment, and Gravel/Cobble). It was observed that the accuracies of the produced habitat maps were reasonable considering the existing challenges, such as the error of the LiDAR data and lacking enough in situ samples for some of the classes such as macroalgae. The overall classification accuracies for the two-class and five-class maps were 87% and 80%, respectively, indicating the high capability of the developed machine learning model for future marine habitat mapping studies. The results also showed that Eelgrass, Fine Sediment, Gravel/Cobble, Macroalgae, and Rockweed cover 22.4% (3.66 km 2 ), 51.4% (8.39 km 2 ), 13.5% (2.21 km 2 ), 6.9% (1.12 km 2 ), and 5.8% (0.95 km 2 ) of the study area, respectively.

Similar Items